Doing Politics
Spies as Informants: Triangulation and
the Interpretation of Elite Interview
Data in the Study of the Intelligence
and Security Services
Philip H.J. Davies
University of Reading
This article examines the application of ‘triangulation’ to the use of elite interviewing in political
science, with specific reference to the study of the intelligence and security services. It is argued
that the problems involved in using elite interviews in security and intelligence studies are no
different than in other areas of political science, but simply more pronounced. It is further argued
that these problems can be most effectively addressed in terms of the sociological ‘triangulation’
strategy of multi-methodological research. The article concludes that this approach is, moreover,
generally applicable to political studies at large.
Introduction
In his 1996 article in Politics on doing elite interviews, David Richards winds up
with the intellectual cliffhanger that, on completing his programme of interviews,
‘the interviewer has now reached one of the most important stages – analysing the
information provided by the interviewees. This, in itself, is no small task’ (Richards,
1996, pp. 203–204). It is indeed no small task, and the problems and options in
handling interview data, elite or otherwise, have been a major issue in sociological
discussion for decades. For the most part, however, the solutions and options
developed in sociological literature have had only the most limited dissemination
outside that field. This is regrettable in view of the widespread concern with the
issue in almost very branch of the social sciences. There is, however, one field of
political studies in which the issues of elite interviewing are particularly pronounced,
and that is the study of intelligence and security agencies. In the following discussion,
I propose to show how the application of certain sociological concepts to the
problems of elite interviewing applied to security and intelligence studies can
provide tools of more general use to political scientists, even where their research
interest is less sensitive and fraught with secrecy than studying espionage.
Information about the intelligence services is notoriously sparse (undoubtedly a
major reason that security and intelligence studies remains a minority interest in
political science), and the official documentary records available at sites such as the
Public Record Office and India Office Library of Records have been ‘sanitised’, with
so-called ‘weeders’ trying to remove any trace of British intelligence information
POLITICS: 2001 VOL 21(1), 73–80
© Political Studies Association, 2001.
Published by Blackwell Publishers, 108 Cowley Road, Oxford OX4 1JF, UK and 350 Main Street, Malden, MA 02148, USA
from available archives. On the whole, the ‘weeders’ have proven a lot less
successful than Her Majesty’s Government might have wished (Andrew, 1984,
pp. 43–65), and since the Major administration’s 1993 ‘open government’
initiative, the quality of the documentary record available has improved dramatically.
However, it continues to be necessary to substitute interviews for official
documents in many areas, particularly concerning the post-1946 history of
Britain’s intelligence machinery. Contemporary analytical studies such as Peter
Gill’s Policing Politics (1994) and Ian Leigh and Lawrence Lustgarten’s In From the
Cold (1994) both rely extensively on interviews to enrich and interpret the arcane
and often vaguely worded intelligence reviews, inquiries and legislation (Leigh and
Lustgarten in particular depend upon information from then-Intelligence Coordinator
Sir Gerald Warner in their ‘Coda on the 1994 Intelligence Services Act’
(1994, pp. 493–516)), while Christopher Andrew’s massively researched Secret
Service (1987) frequently falls back on that unfortunately opaque convention of
historical writing, ‘private information’. A recent project on the institutional
development on the Secret Intelligence Service (MI6) included a programme of
interviews with 13 former SIS personnel, interviewed in depth concerning organisation
and management in that agency (Davies, 1997 and 2000), and in this
context it proved necessary to develop some sort of a systematic approach to correlating
their information with other available sources, and each other.
The Inevitability of Interviews
Secrecy and secret services do not, however, represent a unique case. Rather, they
bring into sharp relief endemic problems in the pursuit of political science in
general, and more specifically in the United Kingdom. The British government is
one of the most secretive democratic governments, and the notorious section 3(4)
of the Public Records Act, retaining documents indefinitely, affects many other
areas besides the study of intelligence and security services. Moreover, the intrinsic
limitations of archival sources mitigate against complete reliance on documentary
sources in almost every area of social-scientific research. Although the clarity and
what might be termed in Pentagonese the ‘survivability’ of documentary records
make them in many respects the preferred sources of historical information, documentary
sources are by no means unassailable. The limitations of the British public
archives have been summarised by the historians Sean Glynn and Alan Booth who
note three problems particularly relevant here: (1) Cabinet papers are incomplete
and thereby potentially misleading; (2) they draw one’s attention to the formal
‘administrative process’ of policy-making rather than the substantive causes and
effects of that policy; (3) finally, much like memoirs, Cabinet, departmental and
‘other political papers’ tend to have what Glynn and Booth call a ‘self-justificatory
element’ (1979, p. 315). The authors even note that at least one case exists where
a participant in the Cabinet process has claimed on the basis of his first-hand
experience of the meetings in question that the official record was intentionally
misrepresentative out of political expediency (Richard Crossman contested the
accuracy of state papers in his memoirs on a number of occasions (Glynn and
Booth, 1979, p. 305)). They conclude, therefore, that ‘There is a real need for
historians to check evidence from the State papers against whatever alternative
sources are available’ (Glynn and Booth, 1979, p. 315). What the authors do not
PHIL H. J. DAVIES74
explicitly deal with is that, on the whole, the minutes of committee meetings such
as the Cabinet or Committee of Imperial Defence often tend to mask over the
political processes of debate and discussion and record only that which was or
could be agreed upon. This is particularly true of the papers of the Joint Intelligence
Committee, a body with an explicit ethos of seeking joint consensus (Herman,
1995, pp. 23–24). Such a limited record can potentially conceal a multitude of sins.
The problems of archival incompleteness have also been noted by Anthony Seldon
and Joanna Papworth in their By Word of Mouth: ‘Elite’ Oral History (1983),
concluding that interviews are in fact a necessary supplement to documentary
records. Indeed, this is much the same rationale for interviews provided by
Richards (1996, p. 100).
What Glynn and Booth are advocating in items (1) and (3) above, and Seldon and
Papworth throughout their volume, points in the direction of the historiographic
equivalent of the sociological device of multi-methodological ‘triangulation’ in
social science research. The triangulation concept is, in its original conception, a
counter to the limitations of the survey-based research which was the prevalent
form of social science research methodology during the 1960s (Webb et al., 1966,
pp. 12–27). In their original formulation of the triangulation strategy, Webb et al.
do indeed discuss the use of official archives, but their goal is gleaning quantitative
sociological rather than qualitative sociological information from archives (1966,
ch.3 and ch.4). The goal is, however, to provide a cross-reference between interview
or questionnaire data and archival records. Webb et al. argue that ‘the most
fertile search for validity comes from a combined series of measures, each with its
own idiosyncratic weaknesses, each pointed to a single hypothesis’ (1966, p. 174).
They conclude, therefore, that ‘with studies of social change, the most practical
method is to rely on available records, supplemented by verbal recall’ (ibid., p. 179).
The goal of triangulation is, so to speak, to provide a parallax view upon events.
A complication exists in the combination of oral and archival evidence in compiling
a qualitative developmental account. That is, the use of different standpoints for
qualitative perception rather than quantitative measurement means that one not
only sees the same thing from a different angle, one sees entirely different facets of
that thing. Thus the point of view in one particular document may reveal entirely
different features of some historical item from other documents, or of a witness or
group of witnesses, and they in turn may provide information additional to that in
the documentary record. Thus while the arguments of Webb et al. in some degree
parallel those of Booth and Glynn, or Seldon and Papworth, the fundamental
difference exists that Webb’s use of multiple methodologies is purely corroborative,
while that employed by the historian is additive as well as corroborative. Due to the
very fragmentary nature of the archival record on the intelligence community, it is
particularly necessary to adopt the combined approach, in which interview
evidence can provide information additional to documentary materials, as well as
corroborative information. However, precisely because of that need to employ oral
evidence, it is also crucial to provide a process of interpretation, collation and
analysis more explicit and transparent than cursory references to ‘senior officials’
or ‘private information’. As a consequence, any such strategy of interpretation,
collation and analysis will very extensively drive or determine the operational
design of the interview programme itself.
DOING POLITICS 75
Elite Interviewing in Security and Intelligence Studies
Elite interviewing is a central tool in the study of intelligence and security services,
not least because intelligence is something created by, of and for the policy-forming
and decision-making elites in national government. The study of SIS referred to
above, examining the structural, institutional development of the SIS from its
inception until the middle decades of the Cold War, inevitably referred to a more
narrowly defined ‘elite’ than is usually the case, specifically former SIS personnel
who could help assemble both a series of time-frame pictures of structure and
process in the SIS, and a sequential sense of how that structure and process evolved.
Because of the very secrecy surrounding such people and their work, assembling a
set of interviewees depended very heavily on a strategy of ‘snowballing’, or asking
each interviewee to recommend one or more other potential contacts. Where
possible, respondents were interviewed on two or more occasions, either in person
or by telephone, as well as written correspondence. This was proved necessary for
two main reasons. In the first place, once an interview was completed and the
content transcribed, uncertainties and ambiguities in the information would
become apparent. In the second place, as will be discussed in greater detail below,
new and unexpected information almost always appeared, creating a need for a
subsequent interview with the respondent providing that information, and a need
to go back to earlier respondents as additional sources of verification and elaboration.
On completing transcription of the interviews, copies of the transcripts were
forwarded to the respondents to try to ensure that the representation of the
respondents’ information was as accurate as possible.
Because of the ages of the former officers interviewed, the interviews themselves
were less an exercise in ‘guided conversation’ than in ‘guided reminiscence’. In
general, interviews in sociology are divided into structured and unstructured types.
Structured interviews are conducted on the basis of a strictly worded and
sequenced series of questions. Unstructured interviews, sometimes referred to as
‘in-depth’ interviews or ‘guided conversations’ rely less on a fixed schedule than a
series of topics to be covered and/or prompts intended to direct the respondent in
particular directions of interest to the researcher(s). Structuring is something of a
matter of degree, and so there is a broad, grey area of ‘semi-structured’ interview
strategies in which the wording or order of the questions on the schedule can be
altered in real time by the researcher conducting the interview. The research design
discussed here employed a semi-structured method, one that employed both
relatively highly structured and relatively unstructured components at different
points. To start with, when going in to talk to someone about their career, it is
crucial to know what that career was and, if possible, its most important features
relevant to the interview. Intelligence officer careers are, however, endemically
secret things and this therefore posed an initial hurdle in schedule design. One
could not very well go into the interview with a vast ‘shopping list’ of requirements
and queries, most of which would not be relevant to the informant, not merely
because of the wasted time but also because the sense of frustration coupled to that
inefficiency could easily undercut any attempt to develop a working rapport.
Consequently, each interview schedule was composed of three distinct stages – a
relatively unstructured initial phases followed by two more highly structured
PHIL H. J. DAVIES76
stages. The unstructured initial phase invited respondents to give a life-account of
their career, from which it was possible to select those questions from the generalpurpose
shopping list that the respondent could address. The third, structured stage
occurred in the event of subsequent interviews, and was a supplementary shopping
list intended to draw out additional information about any new items of information
that emerged during the previous two stages. Once completed, there was also
a need to find a way to refer to the interview data as explicitly as possible while
retaining the confidentiality of the informants, but without resorting to the vagueness
and opacity of conventions such as ‘private information’. The solution to this
problem was also drawn from sociological methodology (which very commonly
delves into sensitive and confidential matters such as sexuality, the family and
healthcare), and that was to assign to each informant a serial number and refer to
information from that informant in terms of that serial number, for example
‘Information from I–28’.
Triangulation and the Evaluation of Elite Interview Data
The epistemological status of information elicited through qualitative interviewing
is, necessarily, another fraught matter. In general, sociological views of qualitative
interview data tend to be cast in terms of the exploration of respondents’
perceptions and sensibilities rather than the factual accuracy of those perceptions
(see, for example, Mason, 1998, pp. 38–42). This mildly relativist position is one
aspect of sociological methodological thought which does not transpose easily to
historical and political studies. Qualitative historical interviewing is really about
eliciting first-hand accounts, securing the evidence of witnesses, as it were (with
the inevitable uncertainties such a metaphor implies). Thus the epistemological
and methodological status of a historical interview is more akin to that of the
memoir than anything else (see, for example, Gamble, 1994; also, with specific
reference to the study of intelligence and security services, the debate between D.C.
Watt and Lawrence Lustgarten in Watt, 1988 and 1989 and Lustgarten, 1989).
That is, for most intents and purposes, memoirs are usable primary sources where
the authors writes from first-hand experience; second-hand and hearsay
information rapidly becomes increasingly unreliable the further the author was
from the event. For the most part, these criteria of reliability are as applicable to
qualitative historical interviews as they are to memoirs. There are also certain
trade-offs between interview and memoir. In the case of the interview, the data
will have a higher level of validity than the memoir (that is, having been collected
in accordance with the specific needs of the project, it will have a better ‘fit’ to
those needs), but the memoir as a published artefact is less volatile over time than
memory, and hence possibly marginally more reliable.1
The next problem is how to weight interview data in with, or possibly against,
other classes of data in the assessment process. Three main criteria can generally be
adopted for the additive use of oral history contributions: first, like the use of
memoirs, the information has to be first-hand reporting; second is the level of
access – a head of station reporting his unit’s organisation would, for the most part,
be viewed as more reliable than a junior officer who was doing the station legwork
at the time; and thirdly is what might be called the informant’s ‘track record’, their
overall displayed reliability where the informant’s recollections appear generally
DOING POLITICS 77
sound when checked against other available sources. Seldon and Papworth
advocate much the same criteria, and draw an analogy between processing oral
history and the rules of witness evidence in English common law (Seldon and
Papworth, 1983, pp. 125–129). However, alternative primary sources are not
always available and in a number of cases interview data may be the only form of
data available, or interview data coupled to information from secondary sources.
Secondary sources have to be treated with considerable caution, not least because
most of the published literature on intelligence has been based almost entirely on
interviews, typically by journalists, with very little in the way of a rigorous
methodology for collection, interpretation and evaluation.
Given the necessary reliance of intelligence studies on interviews and memoirs,
arbitrarily mixing and matching the versions of events emerging from reported
evidence is simply not a viable approach. The optimum solution appears to be a
triangulation triad of primary sources (interviews, published first-hand accounts;
and documentary sources (published or archival)), with published secondary-source
information available in cautious reserve (see Figure 1). Information can then be
cross-referenced both between and within the data types employed.
The most difficult aspect of this kind of triangulation process is not the process of
corroboration through multiple sources, but the incorporation of additional
information from first-hand accounts. Where interviews alone are the available
source for a particular item, a good practice is that adopted by Nicholas Eftimiades
in his study of Chinese foreign human intelligence operations; that is, a minimum
of two independent interview sources are required for any item to be treated with
any real confidence (Eftimiades, 1994, p. 11). The argument in this latter case is,
essentially, a variation on the old information-theory strategy of improving the
signal to noise ratio of a message by sending the same message down two
redundant channels. That is, the noise on each ‘channel’ (i.e. each informant) is
particular to that channel or individual but the signal is common to both, and so
what they have in common is more likely to be signal than noise. In the SIS study,
this was taken as a minimum criterion for the addition of any information from
interview sources on their own, while interview data might be added to documentary
information only when there was already evidence of consistency between
Figure 1: Triangulation sources
PHIL H. J. DAVIES78
Interview Interview
Memoirs Documents
Secondary
Sources
the two, i.e. indicating the probable reliability of the interview data on the subject
at hand.2
Where interviews came into conflict with other interviews or memoirs, if
the item was small enough to be passed over in silence on the grounds that the
results were unreliable, this was the preferred course. Where interviews or memoirs
conflicted with written records, for the most part the written records were taken as
the final authority. Where the item has proven too significant to be dropped from
the historical narrative, then two options existed. In the first place, where a clear
consensus, albeit not unanimity, was displayed, the majority view was adopted,
although the uncertainty is reported in the source notes. Where a complete lack of
consensus existed, and no documentary sources were available to tip the scales one
way or the other, the practice was to report on the uncertainty in the text. In
the last analysis, the basic rationale for an explicit strategy of triangulation is that
where there exists such a palpable degree of uncertainty – as is endemic in the study
of intelligence – if a particular version of events is to be adopted and presented as
probably accurate (particularly in preference to alternative versions) there must be
strong, clearly defined and articulated criteria for making that judgement.
Conclusion
Nine years ago, it could indeed be argued that what intelligence studies needed was
not a combination of speculation and bemoaning the incompleteness of the documentary
record, but rather ‘better use … of what records do exist and of systematic
methods of enquiry, such as structured interviewing of former intelligence officers’
(Robertson, 1987, p. 98). Such an approach does, indeed, have considerable
potential when combined with and added to alternative sources in a strategy of
triangulation, whether the triangulation strategy is corroborative, additive or both.
Perhaps more significantly, given the pervasive secrecy of the British government
at large, such a strategy of triangulation is inevitably of far more general use than
simply the study of so-called ‘secret services’. Government secrecy is a major
problem for anyone doing any branch of political studies in the United Kingdom,
while intrinsic difficulties with other sources, such as self-serving biases in published
sources and unavoidable incompleteness in documentary archives, mean that
the need for qualitative elite interviewing is a general one, and therefore the need
for a systematic approach to evaluating that data and integrating it with information
from different types sources is also a general one.
Notes
1 There also exists a peculiar, marginal category of memoir where the author has fortified their personal
recollections with reference to documents in their possession. In the study of the British Secret
Intelligence Service, for example, there are three particularly notable examples: Compton
Mackenzie’s three volumes of memoirs (1931, 1939 and 1940) make extensive use of surviving
correspondence in Mackenzie’s possession 20 years after the events recounted; R.V. Jones’s Most
Secret War (1978) was written on the basis of an extensive private holding of surviving intelligencerelated
materials, while more recently Nigel Clive employed similar papers to reproduce political
intelligence reports submitted to SIS Far East and detailed mule-and-agent expenses of running his
late wartime Greek intelligence networks in his Greek Experience (1985). A somewhat ambiguous
case is accounts such as J.G. Beevor’s SOE: Recollections and Reflections 1940–45 (1981), written
with the aid of document summaries released to him by the Foreign Office. Similarly, during my
interviews, a couple of respondents referred to documents in their possession or to which they had
access. Much as one would treat memoirs of the type described as being more reliable than accounts
drawn purely from memory, so this type of interview data has to be given greater weight than pure
recollection.
DOING POLITICS 79
2 Just because a source is reliable on one matter does not guarantee reliability on everything about
which they provide an account.
References
Aldrich, R.J. (1997), ‘SIS in the Far East’, Modern Asian Studies, pp. 179–217.
Andrew, C. (1984), ‘Historical Research into the British Intelligence Community’ in R. Godson (ed.),
Comparing Foreign Intelligence, London: Pergamon-Brassey, pp. 43–65.
Andrew, C. (1987), Secret Service, London: Sceptre.
Beevor, J.G. (1981), SOE: Recollections and Reflections, London: Bodley Head.
Clive, N. (1985), Greek Experience, Wilton, Salisbury: Michael Russel 1987.
Davies, P.H.J. (1997), ‘Organisational Development of Britain’s Secret Intelligence Service 1909–1979’
(unpublished Ph.D. thesis, Reading).
Davies, P.H.J. (2000), ‘MI6’s Requirements Directorate: Integrating Intelligence into the Machinery of
Government’, Public Administration 78(1), pp. 29–49.
Eftimiades, N. (1994), Chinese Intelligence Operations, London: Frank Cass.
Gamble, A. (1994), ‘Political Memoirs’, Politics 14(1).
Gill, P. (1994), Policing Politics, London: Frank Cass.
Glynn, S. and A. Booth (1979), ‘The Public Records Office and Recent British Economic Historiography’,
The Economic History Review 32(3), pp. 303–315.
Herman, M. (1995), ‘Assessment Machinery: British and American Models’, Intelligence and National
Security 10(4), pp. 13–35.
Jones, R.V. (1978), Most Secret War: British Scientific Intelligence 1939–1945, London: Coronet.
Leigh, I. and L. Lustgarten (1994), In From the Cold: National Security and Parliamentary Democracy, Oxford:
Clarendon Press.
Lustgarten, L. (1989), ‘Learning from Peter Wright: A Response to D.C. Watt’, Political Quarterly 60,
pp. 222–236.
Mackenzie, C. (1931), First Athenian Memories, London: Cassell.
Mackenzie, C. (1939), Greek Memories, London: Cassell.
Mackenzie, C. (1940), Aegean Memories, London: Chatto & Windus.
Mason, J. (1998), Qualitative Researching, London: Sage.
Richards, D. (1996), ‘Elite Interviewing: Approaches and Pitfalls’, Politics 16(3), pp. 199–204.
Robertson, K. G. (1987), ‘An Agenda for Intelligence Research’, Defence Analysis 3(2), pp. 95–101.
Seldon, A. and J. Papworth (1983), By Word of Mouth: ‘Elite’ Oral History, London: Methuen.
Watt, D.C. (1988), ‘Fall-out from Treachery: Peter Wright and the Spycatcher Case’, Political Quarterly 59.
Watt, D.C. (1989), ‘Learning from Peter Wright: A Reply’, Political Quarterly 60, pp. 237–238.
Webb, E.J., D.T. Campbell, R.D. Schwartz and L. Sechrest (1966), Unobtrusive Measures: Nonreactive
Research in the Social Sciences, Chicago: Rand Macnally.
PHIL H. J. DAVIES80